Update README.md

README.md

@@ -17,19 +17,19 @@ datasets:
 # Llama 2 13b Chat German - GGUF
 
 This repository contains [jphme/Llama-2-13b-chat-german](https://huggingface.co/jphme/Llama-2-13b-chat-german) in GGUF format.
-The original model created by [jphme](https://huggingface.co/jphme) and is a fine-tune of [Llama2 13b Chat](https://huggingface.co/meta-llama/Llama-2-13b-chat) from Meta, focused on German instructions (helpful for RAG). 
+The original model was created by [jphme](https://huggingface.co/jphme) and is a fine-tune of [Llama2 13b Chat](https://huggingface.co/meta-llama/Llama-2-13b-chat) from Meta, trained on German instructions. 
 
 
 ## Model Sheet
 
-| **Model Attribute**            | **Details**                                                                                                  |
+| **Model Attribute**      | **Details**                                                                                                  |
 |--------------------------|--------------------------------------------------------------------------------------------------------------|
 | **Format**               | GGUF                                                                                                         |
 | **Converted with**       | llama.cpp (Commit: 9e20231)                                                                                  |
-| **Quantization Levels**  | - 8 Bit <br> - 5_K_M Bit <br> - 4_K_M Bit                                                                    |
+| **Quantization Levels**  | 8 Bit<br> 5 Bit K_M <br> 4 Bit K_M                                                                           |
 | **Model**                | [jphme/Llama-2-13b-chat-german](https://huggingface.co/jphme/Llama-2-13b-chat-german)                        |
 | **Created by**           | [jphme](https://huggingface.co/jphme)                                                                        |
-
+| **Training Data**        | Prorietary German Conversation Dataset, German SQuAD, German legal SQuAD data, augmented with "wrong" contexts, to improve factual RAG |
 
 ## Replication Steps
 Clone and install llama.cpp *(Commit: 9e20231)* and use the provided `convert.py` file to convert the original model to GGUF with FP16 precision. The converted model will then be used to do further quantization.